In the spirit of total elimination, the major change here is theĭisappearance of the headers. Pull requests and bug reports and such are welcome viaĮxecutive Summary: Alas, poor SphinxBase! Transitions (“null” transitions for the less FST-aware), but it The JSGF compiler is back to producing unreasonable numbers of epsilon Things on various emulators, but it’s slow and unwieldy to do that. The last known portability issue (which was actually, like, a bug) isįixed and you won’t get unpredictable and bad results on MIPS systems. The big news is that force-alignmentĪnd subword alignment are now quite doable, from the command-line,įrom the C API, and from Python. Ps_activate_search() to do that, and not ps_set_search(), because The names wereĬhanged because they don’t set anything and you have to actuallyĪctivate the search module afterwards. They have the wrong names, use add_* instead. Though you may get some deprecated warnings when you try to use the Python code should continue to work as before, There are a lot of changes so I suggest you look at the release notesĪt the link above. Yes, it’s that time of week again, time for another release
Pick Decoding Model Based on Context Dependence by Output an error message when we cannot execute a tool by Make an option in config for not folding case in phonemes by Use consistent shebang for python by Add -sox flag to sphinx_fe to convert files with SoX by Update and enable G2P code by Librispeech training template by can download it from the release pageĮxecutive Summary: Please try this one, there won’t be another. G2P support has been updated for the most commonly installed version of OpenFST (do not try to use any other version, because C++, that’s why). It is not necessary to install SphinxTrain system-wide to run training. The -remove_silence option has been disabled by default (unlike in PocketSphinx you can still turn it on if you really want to, it might save you a bit of time in training). There is “continuous integration” now (sort of). The dependency on Autotools is gone (CMake ain’t great but it’s much less bad). The dependency on Visual Studio for buliding on Windows is gone (but please just use WSL, please). Multi-CPU training actually works, tested on up to 64 CPUs with LibriSpeech, much easier than setting up PBS on the Clown. The dependency on Sphinx3 for VTLN and force-alignment is gone ( sphinx3_align is included). 
The dependency on SphinxBase is gone, because SphinxBase is gone. To quote the release notes, this release fixes a few long-standingīugs in SphinxTrain and makes the package (hopefully) easier to use. Still working on the other tutorials, sorry. 
Modeling tutorial has been updated to reflect the There is also an updated release of SphinxTrain, and the acoustic
0 kommentar(er)
